Contextual Thompson Sampling via Generation of Missing Data

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Kelly W., Cai, Tiffany Tianhui, Namkoong, Hongseok, Russo, Daniel
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917074591285248
author Zhang, Kelly W.
Cai, Tiffany Tianhui
Namkoong, Hongseok
Russo, Daniel
author_facet Zhang, Kelly W.
Cai, Tiffany Tianhui
Namkoong, Hongseok
Russo, Daniel
contents We introduce a framework for Thompson sampling (TS) contextual bandit algorithms, in which the algorithm's ability to quantify uncertainty and make decisions depends on the quality of a generative model that is learned offline. Instead of viewing uncertainty in the environment as arising from unobservable latent parameters, our algorithm treats uncertainty as stemming from missing, but potentially observable outcomes (including both future and counterfactual outcomes). If these outcomes were all observed, one could simply make decisions using an "oracle" policy fit on the complete dataset. Inspired by this conceptualization, at each decision-time, our algorithm uses a generative model to probabilistically impute missing outcomes, fits a policy using the imputed complete dataset, and uses that policy to select the next action. We formally show that this algorithm is a generative formulation of TS and establish a state-of-the-art regret bound. Notably, our regret bound depends on the generative model only through the quality of its offline prediction loss, and applies to any method of fitting the "oracle" policy.
format Preprint
id arxiv_https___arxiv_org_abs_2502_07064
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Contextual Thompson Sampling via Generation of Missing Data
Zhang, Kelly W.
Cai, Tiffany Tianhui
Namkoong, Hongseok
Russo, Daniel
Machine Learning
Artificial Intelligence
We introduce a framework for Thompson sampling (TS) contextual bandit algorithms, in which the algorithm's ability to quantify uncertainty and make decisions depends on the quality of a generative model that is learned offline. Instead of viewing uncertainty in the environment as arising from unobservable latent parameters, our algorithm treats uncertainty as stemming from missing, but potentially observable outcomes (including both future and counterfactual outcomes). If these outcomes were all observed, one could simply make decisions using an "oracle" policy fit on the complete dataset. Inspired by this conceptualization, at each decision-time, our algorithm uses a generative model to probabilistically impute missing outcomes, fits a policy using the imputed complete dataset, and uses that policy to select the next action. We formally show that this algorithm is a generative formulation of TS and establish a state-of-the-art regret bound. Notably, our regret bound depends on the generative model only through the quality of its offline prediction loss, and applies to any method of fitting the "oracle" policy.
title Contextual Thompson Sampling via Generation of Missing Data
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2502.07064